Static relocation of functions in dynamic libraries must use the PLT
entry as the target. Before this commit, it used to be done in 2 parts
for ARM, with the offset of the PLT entry from the beginning of the PLT
being put in the relocated place in build_got_entries () and then the
address of the PLT being added in relocate_section.
This led to code dealing with reading the offset of a bl instruction in
build_got_entries. Furthermore, the addition of the address of the start
of the PLT was done based on the relocation type which does not convey
whether a PLT entry should be used to reach the symbol.
This commit moves the decision to use the PLT as the target in
relocate_section, therefore having the instruction aware code contained
to the target-specific bit of that function (in <target>-link.c).
Note that relocate_syms is *not* the right place to do this because two
different relocations for the same symbol can make different decision.
This is the case in tcc -run mode where the static and dynamic
relocation are done by tcc.
Storing the PLT entry address in the symbol's st_value field and relying
on the specific relocation type being used for dynamic relocation would
work but the PLT entry address would then appear in the static symbol
table (symtab). This would also make the static symbol table entry
differ from the dynamic symbol table entry.
Change alloc_sym_attr into get_sym_attr and add a parameter to control
whether to allocate a new symattr structure or return NULL if symbol is
not found;
Currently GOT/PLT creation happens in two locations depending on whether
the GOT/PLT [entry] is required by the symbol or the relocation:
- bind_exe_dynsym for relocations to undefined symbol
- build_got_entries/put_got_entry for relocations that require a GOT/PLT
entry
This commit consolidate GOT/PLT creation in build_got_entries by
reducing bind_exe_dynsym's job to create a dynamic symbol for undefined
symbols. build_got_entries then invoke put_got_entry if the symbol being
relocated is undefined or the relocation asks for a PLT or GOT [entry].
put_got_entry is also modified to only export a symbol in the dynamic
symbol table when we are in the case of PLT/GOT [entry] required by the
relocation (since undefined symbol are already exported by
bind_exe_dynsym).
Currently we always build a GOT when we recognize a relocation in
build_got_entries even if the relocation does not require one. In the
same spirit, when the relocation does require one we always create a GOT
entry even if not entry is necessary. This patch restricts the creation
of a GOT and a GOT entry to relocations that needs it, ie:
- do not create a GOT if relocation is not related to GOT and symbol is
not UNDEF
- do not create a GOT entry if relocation only relates to beginning of
GOT
On ARM targets, the jump to ld.so resolution routine is done in PLT0 by
loading the offset to the GOT found in PLT+16 and from there loading the
address in GOT+8 and jumping to it.
Currently tcc starts the first regular PLT entry at PLT+16 which thus
does not contain the offset to the GOT. This commit fixes that.
Note that calls via PLT still worked nonetheless because of some missing
dynamic tag which makes ld.so behaves as if RTLD_BIND_NOW was specified
in the environment for all executable created by tcc.
add_elf_sym is a confusing name because it is not clear what the
function does compared to put_elf_sym. As a matter of fact, put_elf_sym
also adds a symbol in a symbol table. Besides, "add_elf_sym" fails to
convey that the function can be used to update a symbol (for instance
its value). "set_elf_sym" seems like a more appropriate name: it will
set a symbol to a given set of properties (value, size, etc.) and create
a new one if non exist for that name as one would expect.
Do not create a new symbol in add_elf_sym if a symbol with same properties
(value, size, info, etc.) already exists. This prevents symbols from
being exported twice in the dynamic symbol table.
Prior to this patch, an error would only be given when a library has an
unresolved undefined symbol if there is no undefined reference for the
same symbol in the executable itself. This patch changes the logic to
check both that the executable has the symbol in its static symbol table
*and* that it is defined to decide if the error path should be followed.
R_386_GOT32X can occur in object files assembled by new binutils, and in
particular do appear in glibc startup code (crt*.o). This patch is
modeled after the x86_64 one, handling the new relocation in the same
trivial way.
Prior to this commit TinyCC was exporting symbols defined in programs
only when they resolve an undefined symbol of a library. However, the
expected behavior (see --export-dynamic in GNU ld manpage) is that all
symbols used by libraries and defined by a program should be exported in
dynsym section. This is because symbol resolution search first in
program and then in libraries, thus allowing program symbol to interpose
symbol defined in a library.
int i = i++ causes a segfault because of missing guard. Looking
recursively at all backend functions called from middle end several more
guard appeared to be missing.
As per grischka comment, always output a space after macro concatenation
instead of trying to detect if it's necessary as the current approach
has a huge cost.
For program manipulating argv or arge as pointer with construct such as:
(while *argv++) {
do_something_with_argv;
}
it is necessary to have argv and arge inside a region. This patch create
regions argv and arge) if main is declared with those parameters.
As was pointed out on tinycc-devel, many uses of get_tok_str gives as
second parameter the value NULL. However, that pointer was
unconditionally dereferenced in get_tok_ptr. This commit explicitely add
support for thas case.
Introduce a new attribute to check the existence of a PLT entry for a
given symbol has the presence of an entry for that symbol in the dynsym
section is not proof that a PLT entry exists.
This fixes commit dc8ea93b13.
Some symbol (such as __gmon_start__ but this one does not matter to tcc)
can have both a R_386_GOT32 and R_386_PLT32 relocation. It is thus not
enough to test if a GOT reloc was already done when deciding whether to
return early from put_got_entry.
For ARM target, tcc uses the soft float ABI when not asked to use hard
float ABI. This means machine without a VFP co-processor generate code
that they cannot run. This commit add a warning for such cases.
When bound check is enabled, tcc tries to relocate a call to
__bound_init in _init. This means that relocation (in tcc_add_bcheck)
must be done after libtcc1.a (which countains __bound_init) is loaded
but before crtn.o is loaded as this finalize _init.
Prior to this commit runtime library was compiled according to the host
because of the macro used to detec what architecture to choose. This
commit fixes this by using the TARGET_* macro instead.
Call fill_got_entry unconditionally from fill_got so as to avoid
warnings on !x86-64 architectures. This can be done since this code path
is only followed by x86-64 architecture anyway.
When checking for exact compatibility between types (such as in
__builtin_types_compatible_p) consider the case of default signedness to
be incompatible with both of the explicit signedness for char. That is,
char is incompatible with signed char *and* unsigned char, no matter
what the default signedness for char is.
Result of float to unsigned integer conversion is undefined if float is
negative. This commit take the absolute value of the float before doing
the conversion to unsigned integer and add more float to integer
conversion test.
This improves commit 5cbe03b9c4 by
avoiding a double transfer when the default float ABI is already softfp.
It's also more clean by expliciting that the ABI is simply changed for
runtime ABI functions.
EABI functions to convert an int to a double register take the integer
value in core registers and also give the result in core registers.
It is thus necessary to move the result back to VFP register after the
function call. This only affected integer to double conversion because
integer to float conversion used a VFP instruction to do the conversion
and this obviously left the result in VFP register. Note that the
behavior is left untouched for !EABI as the correct behavior in this
case is unknown to the author of this patch.
- Build libtcc1 for cross-compiler on arm (arm to X cross compilers)
- Install libtcc1 and includes for arm to i386 cross compiler
- Add basic check of cross-compilers (compile ex1.c)
The procedure calling standard for ARM architecture mandate the use of
the base standard for variadic function. Therefore, hgen float aggregate
must be returned via stack when greater than 4 bytes and via core
registers else in case of variadic function.
This patch improve gfunc_sret() to take into account whether the
function is variadic or not and make use of gfunc_sret() return value to
determine whether to pass a structure via stack in gfunc_prolog(). It
also take advantage of knowing if a function is variadic or not move
float result value from VFP register to core register in gfunc_epilog().
Move the logic to do a test of an integer value (ex if (0)) out of
arch-specific code to tccgen.c to avoid code duplication. This also
fixes test of long long value which was only testing the bottom half of
such values on 32 bits architectures.
When statically linking, runtime library should be static as well. tcc
could link with libgcc.a but it's in a gcc version specific directory.
Another solution, followed by this patch, is to use libtcc.a when
statically linking, even if USE_LIBGCC was configured.
Use comisd / fcompp for float comparison (except TOK_EQ and TOK_NE)
instead of ucomisd / fucompp to detect NaN comparison.
Thanks Vincent Lefèvre for the bug report and for also giving the
solution.
Set *palign for VT_BITFIELD and VT_ARRAY types in classify_x86_64_arg as
else you happen to have in *palign what was already there. This can
cause gfunc_call on !PE systems to consider an array as 16 bytes align
and trigger the assert if the previous argument was 16 bytes aligned.
- fix computation of absolute value (clearing the sign bit does not
since integers are encoded in 2's complement)
- test sign of integer in a more conventional way (binary and with the
high bit does not work for long long due to a bug in gtst)
- spacing in include
Thomas Preud'homme